The lysolipid transporter Mfsd2a regulates lipogenesis in the developing brain

Abstract

Brain development requires a massive increase in brain lipogenesis and accretion of the essential omega-3 fatty acid docosahexaenoic acid (DHA). Brain acquisition of DHA is primarily mediated by the transporter Major Facilitator Superfamily Domain containing 2a (Mfsd2a) expressed in the endothelium of the blood-brain barrier (BBB) and other abundant cell types within the brain. Mfsd2a transports DHA and other polyunsaturated fatty acids (PUFAs) esterified to lysophosphatidylcholine (LPC-DHA). However, the function of Mfsd2a and DHA in brain development is incompletely understood. Here, we demonstrate, using vascular endothelial-specific and inducible vascular endothelial-specific deletion of Mfsd2a in mice, that Mfsd2a is uniquely required postnatally at the BBB for normal brain growth and DHA accretion, with DHA deficiency preceding the onset of microcephaly. In Mfsd2a-deficient mouse models, a lipidomic signature was identified that is indicative of increased de novo lipogenesis of PUFAs. Gene expression profiling analysis of these DHA-deficient brains indicated that sterol regulatory-element binding protein (Srebp)-1 and Srebp-2 pathways were highly elevated. Mechanistically, LPC-DHA treatment of primary neural stem cells down-regulated Srebp processing and activation in a Mfsd2a-dependent fashion, resulting in profound effects on phospholipid membrane saturation. In addition, Srebp regulated the expression of Mfsd2a. These data identify LPC-DHA transported by Mfsd2a as a physiological regulator of membrane phospholipid saturation acting in a feedback loop on Srebp activity during brain development.

Fig 1. Mfsd2a is required at the BBB for postnatal brain growth.

(A) Reduced brain weights of 8-week-old 2aECKO relative to 2a^fl/fl mice. Data are represented as mean ± SE. 2a^fl/fl, n = 4; 2aECKO, n = 4. **p < 0.01. (B) Coronal sections indicate reduced brain size of 8-week-old 2aECKO relative to 2a^fl/fl mice. Sections were stained with Hoechst. Scale bar, 1 mm. (C) Reduced brain weights of 4-week-old 2aiECKO relative to 2a^fl/fl mice. Data are represented as mean ± SE. 2a^fl/fl, n = 8; 2aiECKO, n = 7. ***p < 0.001. (D) Coronal sections indicate reduced brain size of 4-week-old 2aiECKO relative to 2a^fl/fl mice. Sections were stained with Hoechst. Scale bar, 1 mm. (E) Representative images of Golgi-stained hippocampal neurons in brain sections of 3-week-old WT and 2aKO, and 4-week-old 2a^fl/fl and 2aiECKO mice. n = 2 of each genotype. Scale bar = 50 μm. (F) Sholl analysis of Golgi-stained hippocampal neurons in sections of brains from 3-week-old WT and 2aKO, and 4-week-old 2a^fl/fl and 2aiECKO mice. An average of 44 neurons were analyzed for each brain. Data are represented as mean ± SE. n = 2 of each genotype. 2aKO neurons were significantly shorter relative to WT between 30 and 120 μm from the soma. 2aiECKO neurons were significantly shorter relative to 2a^fl/fl between 30 and150 μm from the soma. *p < 0.05; **p < 0.01. (G) Total dendrite length per Golgi-stained hippocampal neuron in sections of brains from 3-week-old WT and 2aKO, and 4-week-old 2a^fl/fl and 2aiECKO mice. An average of 82 neurons were analyzed for each brain. Data are represented as mean ± SE. n = 2 of each genotype. ****p < 0.001. Numerical values underlying panels 1A, C, F, and G can be found in S1 Data.

Fig 2. Mfsd2a expression at the BBB is indispensable for DHA accretion during postnatal brain growth.

(A) Reduced brain weights of P8 2aECKO relative to 2a^fl/fl mice. Data are represented as mean ± SE. 2a^fl/fl, n = 5; 2aECKO, n = 5. **p < 0.01. (B) Targeted lipidomic analysis of brains from P8 2a^fl/fl and 2aECKO mice. Percentage PC, PE, and PS containing DHA, AA, and phospholipids with both fatty acids having a cumulative 3 double bonds are shown. Data are represented as mean ± SE. 2a^fl/fl, n = 5; 2aECKO, n = 5; biological replicates. **p < 0.01. (C) Reduced brain weights of P8 2aiECKO relative to 2a^fl/fl mice. Data are represented as mean ± SE. 2a^fl/fl, n = 5; 2aiECKO, n = 5. **p < 0.01. (D) Targeted lipidomic analysis of brains from P8 2a^fl/fl and 2aiECKO mice. Percentage PC, PE, and PS containing DHA, AA, and phospholipids with both fatty acids having a cumulative 3 double bonds are shown. Data are represented as mean ± SE. Inset contains enlarged scatterplots when differences between groups are obscured by the scale of the y-axis. 2a^fl/fl, n = 4; 2aiECKO, n = 5; biological replicates. *p < 0.05. Experimental data depicted in this figure can be found in S1 Data.

Fig 3. De novo lipogenesis pathways are up-regulated in DHA-deficient brains.

(A) Gene microarray analysis of brains from P8 mice. Heatmap represents commonly up-regulated Reactome pathways with an FDR ≤5%. Only pathways that are commonly up-regulated in at least 2 genotype comparisons (2aKO versus WT or 2a^fl/fl versus 2aECKO/2aiECKO) are shown. RNA was pooled from 5–6 cerebrums according to their genotype. Color key indicates negative logarithm of the FDR for the selected pathways, for which a higher value indicates the pathway is more significantly different. (B) Increased Srebp-1 pathway genes in Mfsd2a deficiency mouse models. Data are represented as MA plots of gene microarray analysis of brains from P8 mice. MA plot is used to visualize intensity-dependent ratio of raw gene microarray data and indicated that the top-ranked pathway that was up-regulated in 2aKO, 2aECKO, and 2aiECKO was the metabolism of lipids and lipoproteins pathway, which contains lipogenic targets of Srebf1 (Srebp-1). Red dots represent individual targets with significant changes in this pathway over the background of expressed genes. (C) Increased Srebp-2 pathway genes in Mfsd2a deficiency mouse models. MA plot of gene microarray analysis of brains from P8 mice. The second highest ranked pathway that was up-regulated in 2aKO and 2aECKO was the cholesterol biosynthesis and steroid metabolism pathway, which contains cholesterogenic targets of Srebf2 (Srebp-2). Red dots represent individual targets with significant changes in this pathway over the background of expressed genes. (D) Confirmation of up-regulation of Srebp-1 and Srebp-2 pathways using direct mRNA quantification by Nanostring analysis on brains from P8 2a^fl/fl and 2aECKO mice. Heatmap represents agglomerative hierarchical clustering based on Euclidean distance using normalized counts of Srebp-1 and Srebp-2 gene targets. Color bar indicates z-score transformations on genes using normalized counts. 2a^fl/fl, n = 5; 2aECKO, n = 5; biological replicates. (E) Confirmation of up-regulation of Srebp-1 pathway using direct mRNA quantification by Nanostring analysis on brains from P8 2a^fl/fl and 2aiECKO mice, in which deletion of Mfsd2a in the BBB was induced by daily injections of 4-OHT injections from postnatal day 0 to 3. Heatmap represents agglomerative hierarchical clustering based on Euclidean distance using normalized counts of Srebp-1 gene targets. Color bar indicates z-score transformations on genes using normalized counts. 2a^fl/fl, n = 5; 2aiECKO, n = 5; biological replicates. (F) Nanostring analysis on Srebp-1 and Srebp-2 gene targets from brains from e18.5 2a^fl/fl and 2aECKO mice showed up-regulation of both pathways in 2aECKO relative to 2a^fl/fl. Heatmap represents agglomerative hierarchical clustering based on Euclidean distance using normalized counts of Srebp-1 and Srebp-2 gene targets. Color bar indicates z-score transformations on genes using normalized counts. 2a^fl/fl, n = 5; 2aECKO, n = 5; biological replicates. (G) Western blot analysis of Srebp-1, Srebp-2, Scd1, and Mfsd2a expression in brain lysates of P13 mice. 2aECKO brains had increased nSrebp-1 relative to 2a^fl/fl brains. nSrebp-2 levels were similar between genotypes. Scd1, a readout of Srebp-1 activity, was more abundant in brains of 2aECKO relative to 2a^fl/fl mice. Mfsd2a expression was highly reduced in 2aECKO relative to 2a^fl/fl brains. Quantification of nSrebp-1, nSrebp-2, Scd1, and Mfsd2a is shown. β-actin served as a loading control. Numerical values underlying panel 3G can be found in S1 Data.

Fig 4. LPC-DHA represses de novo lipogenesis pathways.

(A) Gene microarray analysis of NSC^WT and NSC^KO treated with or without 50 μM LPC-DHA. Heatmap represents commonly down-regulated Reactome pathways in response to LPC-DHA treatment relative to untreated control cells with an FDR ≤5%. RNA was pooled from 4–5 biological replicates according to their genotype and treatment condition. Color key indicates negative logarithm of the FDR for the selected pathways, for which a higher value indicates the pathway is more significantly different. (B) Srebp-1 target genes were down-regulated by LPC-DHA treatment. MA plot of gene microarray analysis of NSC^WT and NSC^KO treated with or without 50 μM LPC-DHA. Lipogenic targets of Srebp-1 were down-regulated in LPC-DHA–treated NSC^WT but not NSC^KO cells. Red dots represent individual Srebp-1 targets with significant changes over the background of expressed genes. (C) Srebp-2 target genes were down-regulated by LPC-DHA treatment. MA plot of gene microarray analysis of NSC^WT and NSC^KO treated with or without 50 μM LPC-DHA. Lipogenic targets of Srebp-2 were down-regulated in NSC^WT but not NSC^KO cells with LPC-DHA treatment. Red dots represent individual targets’ significant changes in this pathway over the background of expressed genes. (D) Confirmation of down-regulation of Srebp-1 and Srebp-2 targets using direct mRNA quantification by Nanostring analysis. Heatmap illustrates percentage of normalized counts of NSCs treated with LPC-DHA over untreated control. Down-regulation of Srebp targets were observed in NSC^WT with LPC-DHA treatment. Three biological replicates for each genotype were used and are indicated by capital letters above each lane. (E) Re-expression of WT Mfsd2a but not transporter inactive D96A mutant of Mfsd2a in NSC^KO cells restored sensitivity of cells to down-regulation of Srebp-1 and Srebp-2 pathways by LPC-DHA treatment. NSC^KO cells were transduced with Mfsd2a WT or Mfsd2a D96A adenovirus and treated with or without 50 μM LPC-DHA. Heatmap illustrates percentage of normalized counts NSCs treated with LPC-DHA over untreated control. Two biological replicates (NSC^KO #1 and NSC^KO #2) were used for each adenovirus transduction indicated above the lane in the heatmap. Control are cells not transduced with adenovirus. (F) Time course for down-regulation of Srebp-1 and Srebp-2 target genes in NSC^WT treated with or without 50 μM LPC-DHA over 1–16 hours. Heatmap illustrates percentage of normalized counts of NSC^WT treated with LPC-DHA over respective time point untreated control. Significant repression of Srebp pathways begin as early as 4 hours after LPC-DHA administration. Technical replicates were carried out for each time point. Numerical values underlying panels 4D–F can be found in S1 Data.

Fig 5. LPC-DHA inhibits Srebp activity and de novo lipogenesis and Mfsd2a is regulated by Srebp.

(A) Western blot analysis of Scd1 and Mfsd2a expression in NSC^WT and NSC^KO treated with or without 50μM LPC-DHA. Scd1 expression in NSC^WT cells decreased with LPC-DHA treatment, but to a lesser extent in NSC^KO cells. β-actin served as a loading control. Capital letters above each lane represent biological replicates for each genotype (n = 5 for NSC^WT and NSC^KO). (B) Western blot analysis of Scd1 and Mfsd2a expression in of NSC^WT and NSC^KO treated with or without 50 μM LPC-DHA or 50 μM DHA. DHA works in an Mfsd2a-independent manner to decrease Scd1 expression to a similar extent in NSC^WT and NSC^KO cells. β-actin served as a loading control. Capital letters above each lane represent biological replicates for each genotype (n = 2 for NSC^WT and NSC^KO). (C) Specificity of LPC species for down-regulation of Scd1. Western blot analysis of Scd1 and Mfsd2a expression in NSC^WT treated with or without 50 μM each of LPC18:1, LPC18:2, LPC18:3, LPC20:4, or LPC22:6. Scd1 expression is more greatly reduced with increasing unsaturation of LPC-PUFAs. Mfsd2a expression remained relatively unchanged. β-actin served as a loading control. Technical replicates were carried out for each LPC treatment. (D) Dose-response effect of LPC-DHA on Srebp transcriptional activity. NSC^WT and NSC^KO cells were transfected with pSynSRE and treated with or without increasing concentrations of LPC-DHA, as indicated, or 25-hydroxycholesterol and cholesterol, as a positive control for repression of Srebp activity. The bar chart shows percent luciferase activity to respective untreated control as mean ± SE. LPC-DHA inhibited Srebp transcriptional activity in a concentration-dependent fashion in NSC^WT but not NSC^KO cells. Three technical replicates were carried out for each treatment condition. *p < 0.05; **p < 0.01, ***p < 0.001; ****p < 0.0001. (E) Specificity of LPC species for down-regulation of Srebp transcriptional activity. NSC^WT cells transfected with pSynSRE and treated with or without 25 μM each of LPC18:1, LPC18:2, LPC18:3, LPC20:4, or LPC22:6 or 25-hydroxycholesterol and cholesterol (sterols). The bar chart shows percent luciferase activity to untreated controls as mean ± SE. Sterols inhibit Srebp and were used as a positive control for inhibition of Srebp transcriptional activity. Increased repression of Srebp transcriptional activity relative to control was observed with increasing unsaturation of LPC-PUFAs, consistent with effects observed for down-regulation of Scd1 expression shown in panel (C). The bar chart shows percent luciferase activity to untreated control as mean ± SE. Three technical replicates were carried out for each treatment condition. **p < 0.01, ***p < 0.001. (F) LPC-DHA reduces nuclear Srebp-1 levels. Western blot analysis of Srebp-1 expression in NSC^WT treated with or without 50 μM of LPC-DHA. Activators and inhibitors of Srebp-1 processing and expression were used as controls to benchmark the effect of LPC-DHA. T0901317 (T0), an LXR agonist; GSK2033 (GSK), an LXR antagonist; Mevastatin plus mevalonic acid (Mev) inhibits HMG-CoAR activity; 25-hydroxycholesterol plus cholesterol (25-HC) inhibits Srebp-2 processing. β-actin served as a loading control. Technical replicates were carried out for each treatment condition. (G) LPC-DHA treatment reduced biosynthesis of sterols and fatty acids (FAs). [¹⁴C]-acetate lipogenesis assay was performed on NSC^WT cells. NSC^WT cells were treated with either 10 μM or 50 μM of LPC-DHA for 16 hours followed by supplementing cell media with [¹⁴C]-acetate for 2 hours. Sterols and fatty acids were purified as described in Materials and methods and quantified by scintillation counting. Uptake is expressed as mean percent DPM of untreated cells ± SE. Sterol synthesis was significantly reduced in 10 μM LPC-DHA–treated NSC^WT relative to control (*p < 0.05). Fatty acid synthesis was significantly reduced in 50 μM LPC-DHA–treated NSC^WT relative to control (**p < 0.01). Two to three technical replicates were carried out for each treatment condition. (H) Mfsd2a levels are regulated by Srebp activity. Western blot of Mfsd2a expression in NSC^WT cells treated with or without activators or inhibitors of Srebp-1 and Srebp-2, as described in (F) above. β-actin served as a loading control. Technical replicates were carried out for each treatment condition. (I) LXR agonist treatment of mice enhances brain levels of Mfsd2a. Western blot analysis of Mfsd2a expression in of brain lysates of WT mice injected with 30 mg/kg/day T0901317 or vehicle control for 4 days. β-actin served as a loading control. Control, n = 4; T0901317, n = 4; biological replicates. Experimental data depicted in this figure can be found in S1 Data.

Fig 6. LPC-DHA affects membrane phospholipid saturation.

Targeted lipidomic analysis of NSC^WT and NSC^KO cells treated with or without 50 μM LPC-DHA for 16 hours. (A) NSC^WT cells showed increased DHA in PC, PE, and PS relative to NSC^KO cells. Data are represented as percentage PC, PE, PS, and LPC containing DHA in phospholipids and represented as mean ± SE. (B) The biochemical pathway mediated by Scd1 is illustrated to highlight the lipid products quantified in this panel. Data are represented as percentage PC and PE phospholipids with fatty acid unsaturation of 2 double bonds and represented as mean ± SE. (C) The biochemical pathway mediated by Fads and Elovl enzymes is illustrated to highlight the lipid products quantified in this panel. Data are represented as percentage of PC, PE, and PS phospholipids containing a total fatty acid unsaturation of 3 double bonds and arachidonic acid (AA) and represented as mean ± SE. (D) Model for physiological role of LPC-DHA in regulating membrane phospholipid homeostasis and brain growth. In the developing brain, de novo lipogenesis mediated by Srebp-1 and Srebp-2 and accretion of DHA in the form of lysophosphatidylcholine (LPC-DHA) uptake at the BBB by Mfsd2a are essential for brain development. LPCs are synthesized by the liver and circulate in blood bound to albumin. In the left panel, Mfsd2a is present at the BBB to transport LPC-DHA from blood into brain, resulting in enrichment of DHA in phospholipid pools during pre- and postnatal brain development. Phospholipid DHA pools attenuate Srebp activity, leading to decreased lipogenesis. Moreover, Mfsd2a itself is regulated by Srebp, forming a negative feedback loop to balance de novo lipogenesis with exogenous uptake of LPC-DHA, with the result of maintaining homeostasis of membrane phospholipid composition. Conversely, in the right panel, Mfsd2a deficiency at the BBB, consistent with the lack of LPC-DHA transport across the BBB into brain, LPC-DHA accumulates in plasma and postnatal brain growth and phospholipid DHA pools are reduced, resulting in microcephaly. Consequential to reduced levels of phospholipid DHA pools prior to the onset of microcephaly, Srebp activity is enhanced, leading to a compensatory increase in lipogenesis. For panels (A–C), 3 biological replicates of NSC^WT control, NSC^WT plus LPC-DHA, NSC^KO control, and NSC^KO plus LPC-DHA were used. ****p < 0.0001; ***p < 0.0002; **p = 0.0021; *p = 0.0332. Numerical values underlying panels 6A–C can be found in S1 Data.